This week Target turned on frictionless checkout inside Google's Gemini app and AI Mode — guests can now go from a conversation to a completed Target purchase without leaving the assistant. Target's own fact sheet, updated June 18, frames it plainly: it's the first mass retailer live across Google AI Mode, Microsoft Copilot, and OpenAI's ChatGPT, and it's a co-developer of the Universal Commerce Protocol — the open standard the checkout runs on.

The same week, our crawler crossed a milestone of its own: 10,000 domains under continuous monitoring. Those two facts belong in the same sentence. One is the demand side of agentic commerce switching on at one of the largest retailers in the US; the other is the supply side we measure, at the scale where "is this real yet?" stops being a fair question.

#10,093 domains, measured every day

As of June 19, the UCP Checker crawler tracks 10,093 domains and re-checks every one of them at least every 24 hours. The current census:

• 10,093 domains monitored — up from ~9,250 when we filed the June state-of-the-ecosystem report less than a week ago
• 8,941 verified UCP merchants — stores with a valid manifest, parseable JSON, and the required fields present (~89% of monitored domains)
• 99.6% of verified stores on the latest v2026-04-08 spec
• 82/100 average UCP Score; ~96% of verified stores (8,610) carry an A grade
• 152ms average manifest response, 94% of verified stores answering in under 200ms — the agent-ready latency threshold
• 94/100 ecosystem stability score, with 95.8% of the directory re-checked in the last 24 hours

This is the boring, load-bearing number. Ten thousand domains, graded the same way, refreshed daily, published openly — a neutral meter anyone can point at. The capability shape is the one we've reported all year: the core shopping surface (checkout 99.7%, cart 99.0%, order management 99.0%, catalog 99.0%) is near-universal, while the deep primitives that turn "agent-shoppable" into "agent-payable" — identity linking (17), AP2 mandate (5), buyer consent (5), embedded checkout (2) — are still a rounding error. Target going live in Gemini is exactly the kind of demand-side pressure that eventually moves those numbers.

#We passed 2,000 agent sessions weeks ahead of forecast

Numbers like 10,093 tell you which stores have UCP. They don't tell you whether an agent can actually buy. That's the second vantage — UCP Playground, where real frontier models run end-to-end shopping sessions against real stores.

Back on May 4, when Playground crossed 1,000 sessions, we forecast the next milestone — 2,000+ recorded sessions — for early August. We passed it weeks ago: the June report logged over 2,000 agent sessions (2,089 runs across 232 stores and 16 models) by mid-June, roughly six weeks ahead of where we expected to be. The harness is being exercised harder and faster than we modeled — which is its own signal about how seriously developers are now treating agent checkout as something to test before they ship.

There's a tidy symmetry worth noting. Target reports that AI-driven traffic to its site grew 2,000% in Q1 year-over-year — five times the ~400% growth across retail sites overall. Two thousand percent more shoppers arriving by agent; two thousand agent sessions on our side stress-testing whether the stores they land on can close the sale. The funnel is the same one we keep documenting: agents find products nearly everywhere, build carts often, then stall — only ~7% of agent-reached checkouts completed a purchase in June. Demand is arriving faster than reliability is.

#Why the pairing matters

For most of UCP's first five months, the honest read was "real supply, thin demand." Thousands of verified stores, but the buyers — the assistants actually sending shoppers through agentic checkout — were mostly demos and pilots. Target turning on Gemini checkout, as a named co-developer rather than a logo on a slide, is the demand side becoming concrete at scale.

That's the moment the measurement matters most. When a retailer the size of Target routes real shoppers through agentic checkout, "does this store work for an agent?" stops being a developer's curiosity and becomes a revenue question. A near-universally A-graded directory still converts an agent checkout only ~7% of the time — and the gap between has a manifest and an agent can buy from it is now a gap with money flowing through it.

That's the lane we've built for: an independent, ungated meter for the supply side, and a runtime harness for whether agents can actually transact against it. As the protocol's co-developers switch on demand, our job stays the same — measure it the same way for everyone, owned by no one.

#What to watch

• Who else goes live in Gemini and Copilot. Target is first among mass retailers; the question is how quickly Best Buy, Wayfair, and the rest of the 20-odd UCP co-developers follow into live agentic checkout rather than pilots.
• Whether demand moves the cliff. Identity linking, AP2 mandate, and buyer consent are still single- and double-digit. Real shopper volume through Gemini is the kind of pressure that historically precedes a capability bundling.
• The next monitoring milestone. Ten thousand domains took five months. At June's discovery rate, the back half of the year is where the supply side either keeps compounding or hits the limit of a one-platform directory.

We'll keep the meter running. The next monthly State of Agentic Commerce lands in mid-July.

#Sources

All adoption figures are from the UCP Checker crawler as of June 19, 2026 (every tracked domain re-checked at least every 24 hours). Agent-session figures are from UCP Playground evals. The verified-merchant dataset is published monthly on Hugging Face under CC-BY 4.0; the developer tools, public API, and bulk checker are ungated.

• Target's announcement: About Target's new frictionless checkout in Google Gemini
• Track adoption live: ucpchecker.com/stats
• Browse the directory: ucpchecker.com/directory
• Run a UCP Score: ucpchecker.com/score
• The 1,000-session milestone (and the forecast): UCP Playground at 1,000+ Agent Sessions
• Latest census: The State of Agentic Commerce — June 2026